Intelligent Handover Management Enabled by O-RAN and Deep Reinforcement Learning

The increased number of connected devices and diverse quality of service demands in cellular networks present formidable challenges for efficient traffic management. In this paper, we present a handover control approach to tackle these challenges in 5G networks, integrating the Proximal Policy Optimization (PPO) framework with Reservoir Computing (RC). Our approach, aligned with the O-RAN architecture, offers a near-real-time solution for intelligent handover management, improving cell performance, and enhancing user experience. This renders our solution O-RAN compliant and deployable as an ’xApp’ on the RAN Intelligent Controller (RIC). The introduced algorithm features a sequential state design including user-specific and cell-specific metrics. Harnessing the power of reservoir computing, our solution captures dynamic changes in state sequences while reducing training overhead and enhancing online training efficiency. The efficacy of our algorithm is demonstrated through evaluations on the ns3 platform, proving its superiority in system-level simulations. The experimental results affirm the effectiveness of the introduced algorithm in addressing the dynamic challenges of handover in cellular networks.